image segmentation

kfoynt · Mar 29, 2019 · 02ab8a8 · 02ab8a8
1 parent 6c37e35
commit 02ab8a8
Show file tree

Hide file tree

Showing 22 changed files with 278,021 additions and 142,118 deletions.
diff --git a/localgraphclustering/find_clusters.py b/localgraphclustering/find_clusters.py
diff --git a/notebooks/affinity_kimon.c b/notebooks/affinity_kimon.c
diff --git a/notebooks/affinity_kimon.cpython-37m-darwin.so b/notebooks/affinity_kimon.cpython-37m-darwin.so
diff --git a/notebooks/affinity_kimon.pyx b/notebooks/affinity_kimon.pyx
@@ -104,3 +104,77 @@ def affinity_gray(int n, int p, int q, double r, double var_I, double var_X, dou
                     col.extend([l, k])
                     data.extend([wkl, wkl])
     return data, row, col
+
+def affinity_multiscale(int n, int p, int q, double r, double var_I, double var_X, double threshold, double[:,:,:] sub_img):
+
+    cdef int i, j, k, l
+    cdef double wkl, d, nF, temp1, temp2, temp3
+
+    cdef list row = [], col = [], data = []
+
+    cdef double[:,:] F
+    cdef double[:,:] X
+
+    F = np.zeros([n, 8])
+    X = np.zeros([n, 2])
+
+    for i in range(n):
+        F[i,0] = 0
+        F[i,1] = 0
+        F[i,2] = 0
+        F[i,3] = 0
+        F[i,4] = 0
+        F[i,5] = 0
+        F[i,6] = 0
+        F[i,7] = 0
+        X[i,0] = 0
+        X[i,1] = 0
+
+    # RGB values
+    for i in range(p):
+        for j in range(q):
+            k = i*q + j
+            F[k,0] = sub_img[i, j, 0]
+            F[k,1] = sub_img[i, j, 1]
+            F[k,2] = sub_img[i, j, 2]
+            F[k,3] = sub_img[i, j, 3]
+            F[k,4] = sub_img[i, j, 4]
+            F[k,5] = sub_img[i, j, 5]
+            F[k,6] = sub_img[i, j, 6]
+            F[k,7] = sub_img[i, j, 7]
+            X[k,0] = i
+            X[k,1] = j
+
+    for k in range(n):
+        #if k%100 == 0:
+        #print k
+        for l in range(k+1, n):
+            temp1 = abs(X[k,0] - X[l,0])
+            temp1 = temp1*temp1
+            temp2 = abs(X[k,1] - X[l,1])
+            temp2 = temp2*temp2
+            d = temp1 + temp2
+            if d < r:
+                temp1 = abs(F[k,0] - F[l,0])
+                temp1 = temp1*temp1
+                temp2 = abs(F[k,1] - F[l,1])
+                temp2 = temp2*temp2
+                temp3 = abs(F[k,2] - F[l,2])
+                temp3 = temp3*temp3
+                temp4 = abs(F[k,3] - F[l,3])
+                temp4 = temp4*temp4
+                temp5 = abs(F[k,4] - F[l,4])
+                temp5 = temp5*temp5
+                temp6 = abs(F[k,5] - F[l,5])
+                temp6 = temp6*temp6
+                temp7 = abs(F[k,6] - F[l,6])
+                temp7 = temp7*temp7
+                temp8 = abs(F[k,7] - F[l,7])
+                temp8 = temp8*temp8
+                nF = temp1 + temp2 + temp3 + temp4 + temp5 + temp6 + temp7 + temp8
+                wkl = exp(-nF/var_I - d/var_X)
+                if wkl > threshold:
+                    row.extend([k, l])
+                    col.extend([l, k])
+                    data.extend([wkl, wkl])
+    return data, row, col